|
In recent years, endoscopic imaging has become the major diagnostic approach for early cancer. It is difficult to accurately obtain the information of lesions and components content by using conventional white light imaging (C-WLI). Multispectral imaging techniques such as narrowband imaging are widely used in endoscopic clinical because of their specificity on the surface structure of digestive tract mucosa. However, there is still a lack of multispectral techniques for tissue components specificity. Tissue components such as lipid, hemoglobin are closely related to the generation and development of tumors, but it is difficult to observe the characteristics of lipid and hemoglobin by conventional white light imaging. Therefore, we studied the multispectral endoscopic imaging technique for the analysis of digestive tract mucosal components. Based on the reflectance spectrum characteristics of these tissue components, we determined their specific wavelength, and comprehensively considered the absorption, scattering, oxygenation and others in order to determine the most appropriate optical band in the wavelength range of 365-660nm. Through experimental verification, we choose different narrow-band wavelength combinations for lipids and hemoglobin oxygen saturation respectively to achieve multispectral imaging. In order to simulate the optical properties of digestive tract surface, we made tissue optical phantom. At the same time, we tested and optimized the imaging system and algorithm by the experiments in vitro, and obtained the optimal multispectral image of tissue components, then realize the quantitative detection combined with the content analysis algorithm. Our pilots show that the multispectral imaging system can improve the contrast of endoscopy image, enhance the detail information, achieve high precision detection of tissue components content, and control the error within 10%.
|
